This article describes how Genmab built an automated AWS pipeline for processing whole-slide images (WSIs) in digital pathology, reducing analysis time from hours to under 30 minutes and cutting manual work by 80%.

• Event-driven pipeline triggered by manifest file uploads to S3 using AWS DataSync
• Coordinator Lambda retrieves metadata; worker Lambda copies WSIs to Bronze S3 bucket
• Two inference options: SageMaker Processing for whole slides, Batch Transform for pre-tiled images
• Custom Docker containers with PyTorch and OpenSlide packaged in Amazon ECR
• Horizontal scaling across GPU instances processes hundreds of gigabyte-scale images in parallel
• Results written to Silver S3 bucket with full lineage tracking in DynamoDB
• Least-privilege IAM, encryption at rest/transit, VPC isolation, and CloudWatch auditing
• Automatic error detection, retry logic, and SNS notifications for operators
• Increased throughput from 50 to hundreds of slides daily with near-real-time processing

Genmab's solution demonstrates how managed AWS services enable scalable, secure, and auditable digital pathology workflows without manual intervention.